Designing Delegation in Requirement Gathering: A Socio-technical Study of Product Practitioners and GenAI

Sammanfattning

As Generative AI becomes increasingly embedded in knowledge-intensive product work, product practitioners are shifting from directly producing requirement artifacts toward configuring, reviewing, and governing AI-assisted outputs. This thesis examines how product managers, product owners, business analysts, and related practitioners design delegation and maintain oversight when using GenAI in requirement-related tasks. While prior research has explored AI augmentation, human–AI collaboration, and AI applications in requirements engineering, limited attention has been given to how practitioners preserve contextual judgment, tacit knowledge, and accountability when AI contributes to requirement documentation. Adopting an interpretive qualitative approach, the study draws on six semi-structured interviews with product practitioners in Sweden. The findings show that AI-assisted requirement work is not a process of simple automation, but an iterative socio-technical configuration. Practitioners selectively delegate bounded and reviewable tasks, translate tacit and domain-specific knowledge into prompts, examples, documents, and constraints, and maintain oversight through continuous evaluation, correction, and accountability practices. The thesis contributes to Information Systems research by conceptualizing GenAI-assisted requirement gathering as boundary design, where requirement quality depends on the interaction between delegation, knowledge translation, oversight, and calibrated trust.